Less wide yet faster - ARM A73

krumme

Diamond Member
Oct 9, 2009
5,872
1,463
136
Another great write from Andrei:
http://www.anandtech.com/show/10347/arm-cortex-a73-artemis-unveiled

2 wide, better branchprediction & fetching
On same node and freq:
Aprox 10% better perf for normal use case scenarios
& > 20% better efficiency
Up to 25% size reduction vs. 72

We get a quadcore on 10nm including 2M L2 on aprox 5mm2. or
2 core 1M L2 on 28nm on aprox 6mm2

Wow. Cheap highend perf to the masses and better efficiency at the same time.
 

NTMBK

Diamond Member
Nov 14, 2011
8,677
1,614
126
Wow. Cheap highend perf to the masses and better efficiency at the same time.
I sincerely doubt that any 10nm part is going to be "cheap". They probably had to reduce size to make up for the increased cost per transistor.
 

Andrei.

Senior member
Jan 26, 2015
316
382
136
I sincerely doubt that any 10nm part is going to be "cheap". They probably had to reduce size to make up for the increased cost per transistor.
He's referring to A73 adoption in 28nm mainstream SoCs. There's great potential in that segment.
 

krumme

Diamond Member
Oct 9, 2009
5,872
1,463
136
I sincerely doubt that any 10nm part is going to be "cheap". They probably had to reduce size to make up for the increased cost per transistor.
You get aprox 10.000 quads including l2 from a wafer, so its not the cpu footprint that is going to hurt profit. 10k. Think about it.

But anyways i was referring to 28nm bulk, because who wants to pay 1usd for the cpu part in a midrange phone :).
Look at eg Samsung midrange and lots of Androids out there. Most on quad or octo a53. Now they can get a dual a73 plus eg. quad a35 on 28nm. Extremely efficient, and yet with a performance profile that match Apple A9. How crazy is that.

I just got a expensive Note 5. 14nm fin fet. Now this tech (a35+a73) could deliver the same experience on 28nm bulk.

The business perspective is fantastic and consumer can move away from that slow a53 at more or less same cost. The way ARM achieved the goals was imo surprising but a wellcome move. Fantastic we can get that performance from prediction and fetching, and slimming the pipeline.

If anyone was in doubt who owns the low power low cost market this is a testament to how things should be done.
 

monstercameron

Diamond Member
Feb 12, 2013
3,829
1
0
what about 22nm fdsoi, that is cheaper than finfet and better than 28nm bulk. I think now is the time for desktop arm to start competing.
 
Last edited:

Andrei.

Senior member
Jan 26, 2015
316
382
136
what about 22nm fdsoi, that is cheaper than finfet and better than 28nm bulk. I think now is the time for desktop arm to start competing.
The ecosystem is not ready for it. There are lots of surrounding IP which are immature and outright missing, such as memory controllers, any analog or RF parts. Once vendors have the ecosystem to build SoCs on FDSOI then it will start being attractive. Currently that's not the case.
 

.vodka

Golden Member
Dec 5, 2014
1,123
1,236
136
Most on quad or octo a53. Now they can get a dual a73 plus eg. quad a35 on 28nm. Extremely efficient, and yet with a performance profile that match Apple A9. How crazy is that.

The business perspective is fantastic and consumer can move away from that slow a53 at more or less same cost.
Yes please. Time for a decent pair of CPU cores to start appearing at the low-mid range instead of a quad A53 or those crazy eight A53 implementations.

I find it interesting they managed to outperform the A72 while reducing complexity and being more efficient with the resources at hand while improving perf/w. Nice engineering.

Let's hope someone can do a cheap yet worthwhile SoC around 2xA73+4xA53 in 28nm that can show up in something along the lines of the Moto G and similar phones.
 
Last edited:

krumme

Diamond Member
Oct 9, 2009
5,872
1,463
136
what about 22nm fdsoi, that is cheaper than finfet and better than 28nm bulk. I think now is the time for desktop arm to start competing.
Was thinking the same. Perhaps Nosta will chime in ;)
Pretty impressive what the french team have done here. A walk from Seronx? :)
 

StrangerGuy

Diamond Member
May 9, 2004
8,428
119
106
He's referring to A73 adoption in 28nm mainstream SoCs. There's great potential in that segment.
Yeah, I get the feeling ARM is positioning the A73 mainly for 28nm since the non-custom ARM licensees can't really make the =<14nm economics work.

My prediction: A73 on cheap-as-dirt 28nm is gonna be so good to the point that big.little and A53 would no longer make sense even for a low-end SoC.
 

el etro

Golden Member
Jul 21, 2013
1,581
14
81
Yeah, I get the feeling ARM is positioning the A73 mainly for 28nm since the non-custom ARM licensees can't really make the =<14nm economics work.

My prediction: A73 on cheap-as-dirt 28nm is gonna be so good to the point that big.little and A53 would no longer make sense even for a low-end SoC.
Makes all sense. But i think that A72 was already clean and cheap enough.

Anyway is pretty elegant extracting such performance of a narrower CPU.



And 10nm process is not looking much good, will come in time, but will not have considerable performance improvements over 14/16nm.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
But anyways i was referring to 28nm bulk, because who wants to pay 1usd for the cpu part in a midrange phone :).
Look at eg Samsung midrange and lots of Androids out there. Most on quad or octo a53. Now they can get a dual a73 plus eg. quad a35 on 28nm. Extremely efficient, and yet with a performance profile that match Apple A9. How crazy is that.

I just got a expensive Note 5. 14nm fin fet. Now this tech (a35+a73) could deliver the same experience on 28nm bulk.

The business perspective is fantastic and consumer can move away from that slow a53 at more or less same cost. The way ARM achieved the goals was imo surprising but a wellcome move. Fantastic we can get that performance from prediction and fetching, and slimming the pipeline.

If anyone was in doubt who owns the low power low cost market this is a testament to how things should be done.
For those low end SoCs (like the Amlogic S905) found in $36 set-top boxes like MXQ Pro this should be really nice.

2.0 Ghz Quad core A53 out and Cortex A73 in.
 

NTMBK

Diamond Member
Nov 14, 2011
8,677
1,614
126
He's referring to A73 adoption in 28nm mainstream SoCs. There's great potential in that segment.
Ah, good point :) Could be interesting for that market, a nice replacement for the A17.
 

videogames101

Diamond Member
Aug 24, 2005
6,768
3
81
Dude what are you on about?

Linux/android and even windows runs on arm.
That might be essentially true, but there are other issues on the desktop side such as widespread device driver support that ARM needs to address. (I would also wager a guess that within 1 or 2 more core generations those issues will be dwindling.)
 
Last edited:

monstercameron

Diamond Member
Feb 12, 2013
3,829
1
0
That might be essentially true, but there are other issues on the desktop side such as widespread device driver support that ARM needs to address. I would also wager a guess that within 1 or 2 more core generations those issues will be dwindling.
The linux/android kernel has tonnes of drivers and this new uarch and new ecosystem probably wouldnt want to support legacy devices.
 
Mar 10, 2006
11,719
2,002
126
Great tech from ARM. A17 evolved into the new mobile high end, A72 line will become their "server" product line beginning with Ares.
 
Mar 10, 2006
11,719
2,002
126
Narrower but higher frequency is an interesting design choice, especially for low-power mobile. Total and complete opposite of Apple's approach (very wide CPU, relatively low Fmax).
 

jhu

Lifer
Oct 10, 1999
11,915
7
76
Narrower but higher frequency is an interesting design choice, especially for low-power mobile. Total and complete opposite of Apple's approach (very wide CPU, relatively low Fmax).
Didn't work so well on desktop. Why are they doing this for mobile?
 

ASK THE COMMUNITY